AG 4   Info

AG Audience

English

60 Minutes

Saarbrücken

In the first part of this work, we focus on affine 3-D reconstruction. Affine reconstruction algorithms may yield results in very short time, since they are solely based on linear algebra. One problem is that either the recovered shapes may be of low quality or the algorithms require parameters to be set in advance. In general, it is not possible to choose these parameters a-priori and the estimated shape may vary greatly, depending on the choice of the parameters. We developed a new affine camera model by introducing one additional parameter. Its main advantage is that we do not need to set any parameter in advance. Rather, their values are automatically determined from the data.

In the second part, we consider the more general perspective camera under unknown and possibly varying intrinsic parameters. Although algorithms which can deal with uncalibrated cameras may not yield 3-D shapes as easily as affine camera algorithms, they are known to have high accuracy. This high precision, however, is obtained in exchange for extremely long computation times. By analytical and empirical means, we determine the causes of this problem and suggest several strategies to greatly reduce computational costs.

Bodo Rosenhahn

--email hidden

passcode not visible

logged in users only